📄 CV Skill Extractor & OpenAI Model Benchmark

Extracts professional skills from a CV (in PDF format) using various OpenAI models and benchmarks their speed, token usage and estimated cost.

This tool provides a simple GUI file picker, processes the CV, queries multiple models, and prints a comparison table.

---

✨ Features

• 📂 Select CV as a PDF via file dialog (Tkinter)
• 🧠 Extract skills using multiple OpenAI models
• ⏱ Measure response time for each model
• 🔢 Count input & output tokens
• 💰 Estimate cost based on model pricing
• 📊 Display benchmark results in a formatted table
• 🧾 Clean bullet-point skill output

---

🧩 Included Models

• gpt-5-mini
• gpt-5-nano
• gpt-4.1-mini
• gpt-4.1-nano

You can modify the list in:

MODELS = [...]

---

💸 Pricing

Pricing is hardcoded (updated 16.02.2026) and used for cost estimation.

PRICING = {
    "model": {
        "input_per_1m": ...,
        "output_per_1m": ...
    }
}

To verify and/or update pricing visit https://developers.openai.com/api/docs/pricing

---

⚙️ Requirements

Install dependencies:

pip install openai pypdf tiktoken tabulate

or

conda install openai pypdf tiktoken tabulate

Tkinter is included with most Python installations.

---

🔑 Environment Setup

Set your OpenAI API key:

export OPENAI_API_KEY="your_api_key_here"

or

setx OPENAI_API_KEY "your_api_key_here"

---

▶️ Usage

Run the script:

python ./cv_parser_benchmark.py

Workflow

1. File dialog opens
2. Select a CV in PDF format
3. Text is extracted
4. Each model:
   • Extracts skills
   • Prints output (to compare with other models)
   • Measures response time
   • Counts tokens
   • Estimates cost
5. Result performance table is displayed

---

📊 Example Output

=== Benchmarking gpt-5-mini ===
Extracted skills:
 • Python
 • SQL
 • Machine Learning
 ...

=== BENCHMARK RESULTS ===

+--------------+----------+--------------+---------------+-------------------+
| Model        | Time (s) | Input Tokens | Output Tokens | Estimated Cost ($)|
+--------------+----------+--------------+---------------+-------------------+
| gpt-5-mini   | 1.32     | 1450         | 120           | 0.00089           |
| gpt-5-nano   | 0.98     | 1450         | 118           | 0.00021           |
+--------------+----------+--------------+---------------+-------------------+

---

🔧 Customization

Change the prompt

Modify:

input = [...]

Benchmark only one model

MODELS = ["gpt-4.1-nano"]

Use your own pricing

Update the PRICING dictionary.

Benchmark more models

Update MODELS and PRICING sections

---

⚠️ Notes

• Token counts are estimated locally (may slightly differ from API billing).
• Very large CVs will increase cost and latency.
• Works with text-based PDFs (not scanned images).
• To work with scanned images it is recommended to execute text extraction on OpenAI side (will increase costs)

---

⭐ If you find this useful, give it a star!